You are on page 1of 42

Chapter 9

Safety Analysis of Hydrogen Vehicles and Infrastructure


Wolfgang Breitung

9.1

Motivation of Safety Investigations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.1.1 Safety-Relevant Properties of Vehicle Fuels . . . . . . . . . . . . . . . . . . . . . . . . . 9.1.2 Statistics of Hydrogen Incidents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2 Analysis of Hydrogen Accidents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.3 Analysis Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.3.1 First Phase: Combustible Mixture Generation . . . . . . . . . . . . . . . . . . . . . . . . 9.3.2 Second Phase: Criteria for Hazard Potential . . . . . . . . . . . . . . . . . . . . . . . . . 9.3.3 Third Phase: Combustion Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.3.4 Fourth Phase: Consequence Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.3.5 Mitigation Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4 Application in the Case of H2 Release in a Garage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4.1 Mixture Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4.2 Hazard Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4.3 Combustion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4.4 Consequence Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.4.5 Summary of Results of Garage Investigations . . . . . . . . . . . . . . . . . . . . . . . . 9.5 Safety Issues for Hydrogen Vehicles and Infrastructure . . . . . . . . . . . . . . . . . . . . . . . . . 9.5.1 Hazards Resulting from Vehicle Operation . . . . . . . . . . . . . . . . . . . . . . . . . . 9.5.2 Safety Issues for Refuelling Stations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.6 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

336 337 341 344 346 347 349 353 354 354 355 356 357 361 363 366 367 368 370 373 375

List of Abbreviations
CAD CFD FZK IRSN Computer-aided design Computational uid dynamics Forschungszentrum Karlsruhe Institute de Radioprotection et de S uret e Nucl eaire

Wolfgang Breitung Institute for Nuclear and Energy Technologies (IKET), Forschungszentrum Karlsruhe, P.O. Box 3640, D-76021 Karlsruhe, Germany, e-mail: breitung@iket.fzk.de

A. L eon (ed.), Hydrogen Technology, c Springer-Verlag Berlin Heidelberg 2008

335

336 KI LEL PRA SL UEL Kurchatov Institute Moscow Lower explosion limit Probabilistic risk analysis Laminar burning velocity Upper explosion limit

W. Breitung

9.1 Motivation of Safety Investigations


Although hydrogen has been safely produced, distributed, and used in chemical industry since many decades, the developed safety procedures and technologies provide only limited guidance for mobile hydrogen applications. In the case of hydrogen-powered vehicles, hydrogen will be used within a decentralised infrastructure in relatively small amounts (several kg per user) by a large population without special training in the safety of combustible gases. The transition in vehicle fuels from liquid hydrocarbons to gaseous hydrogen requires an adaptation of automobile design and safety technology to the special properties of hydrogen. The public will only accept hydrogen transport technology, if a safety level comparable to that of current gasoline vehicles can be obtained. This requires a systematic investigation of the hydrogen behaviour for normal operating conditions, for component mal-functions, for vehicle collisions, and for service/repair conditions. Unacceptable risks should be identied early in time and prevented by design measures before the acceptance of hydrogen technologies is adversely aected. This chapter aims at giving an idea of how a safety analysis can be performed for hydrogen vehicles and infrastructure. The phenomenology of hydrogen-related incidents will be described rst by comparing the safety-relevant properties of hydrogen, methane, propane, and gasoline vapour. It will be obvious that large differences exist between hydrogen and hydrocarbons. Then, the consequences of the specic hydrogen properties will be highlighted by discussing the statistics of hydrogen accidents. As can be supposed, the outcome of H2 -related incidents depends on many parameters and can vary largely, mainly depending on the combustion regime obtained. Therefore, a description of an analysis of hydrogen accidents will follow, which will include the general sequence of events, the analysis procedure, and the mitigation measures. The analysis procedure for hydrogen accidents is based on three-dimensional numerical simulation of hydrogen distribution and dierent hydrogen combustion modes. This analysis methodology shall then be applied to investigate the release of hydrogen from a car parked in a residential garage. It will be demonstrated that mixtures can develop, which allow for fast turbulent deagrations or detonations under certain conditions. To understand the potential hazard caused by such mixtures, local explosion experiments were performed in a closed test chamber simulating the garage situation. The measured (and simulated) overpressure data were then compared to known thresholds for structural damage and human injuries. This example will quantify some possible consequences of H2 release in a conned geometry. Finally, a summary of

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

337

safety issues for hydrogen vehicles and infrastructure in general shall be presented. The potential hazards caused by vehicle operation and infrastructure installations will be discussed. The risk connected with postulated accident scenarios can be assessed by probabilistic risk analysis (PRA) and CFD simulations. The latter concentrates on the risk-dominating sequences identied by the PRA studies. As an outlook, some important research topics will be given, which need to be addressed in future work. There are no indications of fundamental safety issues which might prevent the safe use of hydrogen in mobile applications.

9.1.1 Safety-Relevant Properties of Vehicle Fuels


The safety standards of modern vehicles using hydrocarbons are excellent and the question arises whether the use of hydrogen could have an impact on safety records and statistics. Are there large changes to be expected? Figure 9.1 compares several important safety-relevant physical and chemical properties of hydrogen and gaseous hydrocarbons like density, buoyancy, etc. . . . As can be seen, signicant dierences exist in the properties of hydrogen and hydrocarbons. Hydrogen exhibits a very low density and ignition energy compared to gaseous hydrocarbons, while it has high values for buoyancy, diusion in air, range of combustible H2 -air mixtures, specic heat of combustion, laminar burning velocity, and detonation sensitivity. Clearly, these properties of hydrogen can inuence the initiating event, the development, and the outcome of an accidental hydrogen release. In the following sections, the expected eects of the specic hydrogen properties will therefore be discussed in a qualitative way. Figure 9.1b displays the buoyancy of hydrogen in air. The large buoyancy of hydrogen compared to gaseous hydrocarbons suggests that if space is available in vertical direction, a mixing with air will be without any consequence in many cases. Indeed, a free cloud of hydrogen in air experiences an initial vertical acceleration of 13 g, as calculated from Eq. (9.1): z = (air /H2 1)g 13 g (9.1)

Thus, the large hydrogen buoyancy will support fast mixing with air. Figure 9.1c compares the diusion coecient of hydrogen and hydrocarbons. The large value of 0.6 cm2 /s for hydrogen indicates that hydrogen distribution may noticeably occur also in downward direction. The characteristic diusion length of hydrogen in stagnant air is given by Eq. (9.2): x = 2 Dt (9.2) where t is the time and D = 0.6 cm2 /s. Therefore, distances of a few cm will be aected by H2 diusion within 10 s. In the case of hydrogen release into a laminar

338
a
5 air gas (Kg/m3) 4 kg/m3 3 2 1 0 0,7 0,6 0,5 cm2/s 0,3 0,2 0,1 0 mJ 0,4

W. Breitung
b
1,0 0,0 1,0 2,0 3,0

Density

Buoyancy

c
0,35

d Ignition energy
0,3 0,25 0,2 0,15 0,1 0,05 0

Dffusion coefficient

=1

e
80

140 120 100 MJ/kg 80 60 40 20 0

f heat of combustion

Flammability limit
UEL-LEL (%) 60 40 20 0

g
3 2,5 2 m/s 1,5 1 0,5 0 0,12

h
0,1 1/mm 0,08 0,06 0,04 0,02

Laminar burning velocity

Detonation sensitivity 1

=1

=1

H2

CH4

C3H8

gasoline vapour

H2

CH4

C3H8

gasoline vapour

Fig. 9.1 Comparison of safety relevant physical and chemical properties of hydrogen, methane, propane, and gasoline vapour

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

339

air ow of constant velocity vo instead of stagnant air, the total transport distance, xtotal , increases by the advection length, vo t, as given by Eq. (9.3) xtotal = 2 Dt + vo t (9.3) Figure 9.1d shows that the spark ignition energy of hydrogen-air mixtures is exceptionally low, about one order of magnitude lower than most of the hydrocarbons [1]. Weaker sparks or other small energy deposition mechanisms may therefore trigger an ignition in hydrogen rather than in hydrocarbons. On the other hand, it is frequently stated that most sparks from static electricity, switches, relays, and electrical motors have energies well above 1 mJ. For such strong ignition sources the dierence in ignition energy does not matter. Other ignition mechanisms, like e.g. gas heating by shock focussing or reection, are very relevant to hydrogen, but not to hydrocarbons. Altogether, a larger variety of potential ignition mechanisms exists for hydrogen than for hydrocarbons. Figure 9.1e displays the concentration range of ammable compositions in air for the dierent gases, depicted as the dierence between the upper explosion limit (UEL) and the lower explosion limit (LEL). Under ambient conditions, the combustible H2 -air mixtures range from about 475 vol.% of H2 . This large range of combustible H2 -air compositions is of high relevance to safety investigations. This means that if a hydrogen leak occurs, a much larger part of the H2 -air cloud can burn out compared to the hydrocarbons case. It should be noted that hydrogen-rich regions of the cloud (> 29.5 vol.% H2 ) cannot react completely due to the local oxygen deciency. However, the partial hydrogen burning generally creates sucient overpressure and expansion ow to mix the residual hot mixture of H2 , steam, and N2 with fresh air. Hence, the remaining hydrogen is consumed in a secondary burning. The heat of combustion of hydrogen on a mass basis (120 MJ/kg) is signicantly higher than that of hydrocarbons due to the small molar weight (Fig. 9.1f). Burning of 1 kg of hydrogen in an accident correspondingly causes more energy to be released into the environment, which may result in higher temperature or pressure loads depending on details of the accident scenario. Figure 9.1g shows that the laminar burning velocity (SL ) of hydrogen-air mixtures is signicantly larger compared to hydrocarbon-air mixtures. SL characterises the general reaction kinetics of the combustion process and it is a fundamental property of a combustible mixture. The high chemical reactivity of H2 -air mixtures in comparison to hydrocarbon-air mixtures could also be observed in turbulent combustion experiments with a partial connement of the burnable mixture [2]. This is illustrated in Fig. 9.2 which summarises the maximum ame speeds measured for stoichiometric hydrogen-, methane-, and propane-air mixtures in an obstructed test tube. In the cylindrical tube surface variable openings exist. Burned and unburned gas could vent radially through these openings. The given vent angle divided by 90 is the fraction of open tube surface available for venting. As can be seen, for methane and propane, small vent angles of only 510 are sucient to suppress ame acceleration completely. This corresponds to 5.511% of the total tube

340 Fig. 9.2 Measured maximum ame speeds in an obstructed tube with vent openings. Hydrogen shows a much more pronounced tendency for ame acceleration than hydrocarbons

W. Breitung

1600 1400

Flame speed, m/s

1200 1000 800 600 400 200 0 0 10 20

30% H2 ( = 1) 25% H2 20% H2 15% H2 9.5% CH4 ( = 1) 4% C3H8 ( = 1) Sound speed 30% H2 in air

30

40

Venting angle , degrees

vent surface

surface. In contrast to this, hydrogen-air mixtures show supersonic ame speeds up to 30 of venting. Moreover, even lean hydrogen-air mixtures with only 15% of hydrogen can burn faster than stoichiometric hydrocarbon-air mixtures. Detonations are supersonic combustion waves in which the unburned gas is ignited by rapid adiabatic compression to a temperature which is higher than the self-ignition temperature. Detonation waves are three-dimensional phenomena which show a cellular structure created by superposition of dierent shock wave fronts. This so-called detonation cell size decreases with increasing chemical reactivity of the mixture, so that 1 can serve as a measure of detonation sensitivity.

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

341

Figure 9.1h displays the detonation sensitivity 1 of the dierent gases. The higher reactivity of the H2 -O2 system results in a higher tendency to undergo spontaneous transitions from a deagration to a detonation compared to hydrocarbons. Deagrations are combustion waves in which the reaction is controlled by diusion of heat and radicals from the burned gas to the unburned gas. In this paper deagrations with ame speeds smaller than or equal to the sound velocity in the unburned mixture will be called slow deagrations, those with higher ame speeds will be referred to as fast deagrations.

9.1.2 Statistics of Hydrogen Incidents


Before going into a deeper analysis of hydrogen safety issues, it is informative to review what happened in the past. What were the causes and the consequences of accidents involving hydrogen? Can general trends be identied? How dierent are the risks from gaseous hydrogen (GH2 ) compared to liquid hydrogen (LH2 )? A comprehensive and recent collection of hydrogen-related incidents can be found in [3]. In this review, 287 accidents with gaseous and liquid hydrogen were analysed and classied in a systematic manner. An older review of hydrogen re and explosion incidents can be found elsewhere [4]. It should be kept in mind that hydrogen accidents are not only a topic of the past, but also of the present. An accident with a compressed hydrogen gas trailer occurred in Germany in June 2005 for example. Figure 9.3 displays the photo of the accident. As can be seen, the hydrogen bottles fortunately remained intact and the hydrogen could be vented safely. From the data collected in [3], the following conclusions can be drawn with respect to the causes, ignition sources, combustion regimes, eect of connement as well as the consequences in a hydrogen-related accident.

Fig. 9.3 Accident with a compressed-hydrogen-gas tanker in Germany, 2005. The hydrogen bottles remained intact and the hydrogen could be vented safely

342

W. Breitung

For about one third of the accidents with GH2 and LH2 , no single cause could be identied. Important causes which have been determined were human error, construction decits, and malfunction of components, both for GH2 and LH2 . The most important ignition sources identied were open re, spontaneous selfignition, electrostatic discharge, and hot surfaces. For GH2 , each of these four ignition sources contributed by 10 2% to the total events, whereas for LH2 , these percentages amount to about half of the given values. The ignition source remained unidentied in about 40% of the cases with GH2 . By contrast, almost 60% of LH2 accidents resulted in no ignition. Figure 9.4 displays the combustion regimes observed in accidents involving GH2 and LH2 , respectively. In more than 90%, accidents with GH2 lead to an ignition and in around 60%, to a fast deagration. With LH2 , almost 60% of the events show no ignition, which subsequently reduces the generation of slow or fast deagrations signicantly. It is evident that the formation of a combustible mixture or the ignition of the resulting cold H2 -air mixture is more dicult to reach with LH2 than with gaseous hydrogen release. The full spectrum of ame speeds from slow deagrations to detonations was observed in hydrogen-related accidents. If the maximum overpressure pmax from combustion is considered,

70 60 Percent of incidents 50 40 30 20 10 0 70 60 Percent of incidents LH2 GH2

Fig. 9.4 Combustion regimes observed in accidents with GH2 and LH2 . With GH2 most cases lead to ignition and fast deagration. With LH2 ignition is much less likely [3]. All combustion regimes of hydrogen ranging from slow deagrations to detonations are observed in accidents

50 40 30 20 10 0 no ignition slow deflagr. fast detonation deflagr.

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

343

( pmax po )/ po 0.02 typically results for a slow deagration of 10 m/s ame speed, and ( pmax po )/ po 20 for a detonation with 2000 m/s ame speed. The possibility of fast deagrations and detonations clearly increases with the degree of connement. Connement prevents venting of burned and unburned gases and supports ame acceleration. Furthermore, turbulence-generating ow obstacles are more likely to occur in rooms than in open space. Important consequences of hydrogen-related accidents are injuries and mortality. Figure 9.5 displays the typical rates from dierent classes of hydrogen accidents which occurred with GH2 and LH2 . As can be seen, events with GH2 lead to signicantly more injuries than with LH2 . The total number of fatalities per incident is comparable for GH2 and LH2 . According to this database, the hydrogen risk is dominated by the release of gaseous hydrogen into the environment or into rooms accessible for persons.

injuries

2,5 affected persons per incident GH2 2 1,5 1 0,5 0

fatalities

injuries

2,5 affected persons per incident LH2 2 1,5 1 0,5 0

fatalities

Fig. 9.5 Injury and mortality rates from dierent classes of hydrogen accidents. Risk is dominated by GH2 release and explosion [3]. The term explosion summarizes fast turbulent deagrations and detonations

Release without ignition

Release Release and and fire explosion

No release

344

W. Breitung

The above discussion of safety-relevant properties of hydrogen and the review of past hydrogen accidents shows that the development and the outcome of an accident scenario depends on many dierent parameters. Due to the complex combustion physics of hydrogen-air mixtures, the damage potential may vary by orders of magnitude. Since the whole spectrum of ame speeds is observed in accidents, a comprehensive modelling of hydrogen distribution and combustion phenomena is necessary for predictive accident analyses.

9.2 Analysis of Hydrogen Accidents


In this chapter, the general sequence of events during a hydrogen accident shall be described. Let us consider as an example a vehicle collision in a tunnel involving a car fuelled with liquid hydrogen. If the vacuum of the thermal tank insulation is lost by mechanical damage of the outer tank structure, the LH2 tank inventory will be released in gaseous form into the environment within 1015 minutes [5]. Figure 9.6 displays the numerical simulation of the accident. Herein, it is assumed that the release point is below the trunk. The rst event once hydrogen is released is the mixing of the cold gaseous hydrogen (20 K) with air (300 K). This will result in gas mixtures which are buoyant, rise upwards, and spread along the tunnel ceiling. The hydrogen distribution phase continues until an ignition occurs. Potential ignition sources are e.g. open re, mechanical sparks, electrical sparks, hot surfaces, and electrostatic discharges. For an ignition to occur, a burnable mixture must exist at the ignition location (475 vol.% H2 ). In case of a successful ignition, a slow quasi-laminar deagration will initially propagate from the ignition location to the surrounding mixture with ame speeds in the order of several m/s. The expansion ow of the hot burned gases pushes unburned gas away from the ignition point. In the presence of obstacles, the ow may become turbulent. Depending on the obstacle density, hydrogen concentration, and geometrical connement, the slow deagration will or will not accelerate and transform into a fast turbulent deagration with typically several 100 m/s ame speed. As can be seen in Fig. 9.6, high hydrogen concentrations were only present near the release point (> 10 vol.% H2 ), but no ow obstacles were located in the upper tunnel region. As a consequence, only a slow deagration with less than 10 m/s ame speed was predicted in this case. However, if disadvantageous conditions exist involving high H2 concentrations, high obstacle density, correspondingly high turbulence levels, a large degree of connement, and/or large geometrical dimensions of the burnable cloud, the fast deagration may turn into a detonation. In this case, ame speeds up to 2000 m/s can be reached. Table 9.1 gives a summary of the general accident phases and the corresponding important parameters. In a complete and mechanistic model calculation the eect of all these parameters on the development and outcome of the accident must be predicted.

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

345

160s

500s

900s

0.00 0.10 0.15 0.05 Hydrogen volume fraction

Flame

902s unburned gas

904s

906s

burned gas: T > 800K

Fig. 9.6 Numerical simulation of a tunnel accident involving release of hydrogen from a LH2 tank, mixing with air, and late ignition of the burnable H2 -air cloud [5]

346

W. Breitung

Table 9.1 General sequence of events in hydrogen accidents and potentially important parameters which may have an impact on the accident development and outcome Accident phase - Hydrogen release - Mixing with air - Ignition - Laminar deagration - Fast turbulent deagration Potentially important parameters - Leak location, H2 release rate (g/s), H2 velocity and direction (plume or jet), total H2 mass released - Flow regime (plume or jet), connement, ow obstacles, geometrical scale, time since beginning of release - Location and time of rst ignition (early or late relative to total release period) - H2 concentration, connement and ow obstacles near ignition point - H2 concentration, degree of connement, ow obstacles away from ignition point, geometrical scale available for ame acceleration - H2 concentration, scale of combustible cloud, connement, reecting and focussing walls

- Detonation

9.3 Analysis Procedure


The hydrogen safety-related research activities at FZK concentrated on the development of an analysis procedure allowing for a complete and deterministic modelling of the hydrogen behaviour in accidents. In the following sections, the individual steps of this methodology will be described briey with special emphasis being placed on the input data required and the resulting output. The methodology is based on three-dimensional numerical simulation tools. Figure 9.7 displays the four phases of the analysis procedure, the generation of a combustible mixture, the criteria for the hazard potential, the combustion simu-

COMBUSTIBLE MIXTURE GENERATION Problem geometry Mitigation Scenario Sources Distribution GASFLOW

CRITERIA FOR HAZARD POTENTIAL Flammability y Flame Acceleration y Detonationtransition y

COMBUSTION SIMULATION

CONSEQUENCE ANALYSIS

Slow deflagration FLAME3D Fast turbulent deflagration COM3D Detonation DET3D

Mechanical and thermal loads


Structural response SDO ABAQUS Human effects

GP-Program

Fig. 9.7 FZK procedure for mechanistic analysis of hydrogen behaviour in accidents [7]

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

347

lation, and the consequence analysis. Each of these phases will be described and discussed in the following sections.

9.3.1 First Phase: Combustible Mixture Generation


The rst phase of the analysis addresses the formation of a combustible mixture in a given installation and accident scenario. The important parameters in this phase are the geometry model of the installation, followed by the presence or absence of mitigation devices. Then comes the accident scenario. This step in the analysis is of importance, since it will strongly aect the remainder of the analysis. Finally, the hydrogen source and the mode of hydrogen distribution will be taken into account in this rst phase. 9.3.1.1 Problem Geometry At the beginning of every simulation, a geometry model of the installation must be made on the basis of 2D cuts or 3D-CAD les. From a mathematical point of view, the problem geometry denes the boundary conditions for the solution of the three-dimensional equations of ow mechanics. The construction of a 3D computational grid for a complex installation can be a demanding and time-consuming task. Often, a compromise between the desirable level of detail in the geometry representation and the computational possibilities has to be found. It should be noted that neglected or not adequately modelled ow paths may have a signicant inuence on the calculated ow solution and the resulting H2 distribution. 9.3.1.2 Mitigation In the modelling of a given installation, it has to be decided whether measures for hydrogen control should be included in the numerical simulation or not. To integrate such mitigation devices in the analysis, veried CFD models have to be available in order to obtain a reliable prediction of their eectiveness concerning the reduction of the hydrogen inventory and the subsequent eects on other relevant parameters during the accident. As a consequence, numerical models for potential H2 reduction systems like spark igniters, glow plug igniters, catalytic recombiners were developed, veried in dierent test series, and implemented in the 3D distribution program GASFLOW. (Catalytic recombiners are devices with catalytic surfaces on which hydrogen can recombine with oxygen from the air without creating an open ame. They have been installed in most German nuclear power plants as mitigation measures against hydrogen release in severe accidents.) It should be noted that if further analysis of the hydrogen behaviour should lead to unacceptable consequences for persons or equipment, then the hydrogen mitigation approach has to be improved.

348

W. Breitung

9.3.1.3 Accident Scenario For a given installation with a dened geometry and H2 mitigation system, the next step of the analysis is to decide on the possible type of accident and the sequences which are the most representative cases. This step requires a very good understanding of the investigated system, although some aspects like common mode failure, description of human behaviour, and completeness of the constructed event tree are often dicult to assess. However, it is neither possible nor necessary to examine all sequences with respect to their hydrogen risk. The analysis should cover a small set of boundary (or worst-case) scenarios which in their entity cover all other possible accidents with respect to the potential accident consequences. 9.3.1.4 Hydrogen Sources Experience has shown that an adequate denition of the hydrogen source is very important, since it has a strong inuence on all the following phases of the accident. For a consistent CFD calculation, mass, momentum, and energy of the released hydrogen are required for the complete duration of the accident. Furthermore, position and direction of the source jet have to be dened. 9.3.1.5 Hydrogen Distribution The knowledge of the time-resolved mass, energy, and momentum uxes of the hydrogen source in the next step allows to calculate the transport and mixing of hydrogen with the air in the installation. The result of this simulation will be the time- and space-dependent temperature as well as the composition of the H2 /air mixture and temperature. A large number of interacting physical processes and thermodynamic states have to be simulated with a high spatial resolution to determine gas compositions within a small enough uncertainty band which allows for a meaningful hydrogen combustion simulation. Appropriate goals for this task are the prediction of absolute hydrogen concentrations with an accuracy of a few percent (e.g. 12 2% H2 ). The CFD code GASFLOW has been developed [6] for the modelling of the 3D hydrogen distribution in complex installations. This code has been tested and veried extensively at FZK [7]. The modelling of hydrogen distribution in general requires the description of: 3D uid ow in a complex 3d domain Convective heat transfer between gas and walls Radiative heat transfer Heat conduction into structures Turbulence modelling, and Mitigative measures, e.g. recombiners and igniters

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

349

The distribution calculation will provide time- and space-resolved hydrogen and air concentrations in the installation. The next question which arises concerns the hazard potential of this mixture. Is it ammable? If yes, when and where does the ignition occur? After ignition, will the ame be able to accelerate under the given conditions or even undergo a transition to a detonation? These questions are addressed in the second phase of the analysis, which will be presented in the following section.

9.3.2 Second Phase: Criteria for Hazard Potential


9.3.2.1 Ignition The distribution analysis may result in burnable mixtures for given time and space domains. To initiate the combustion process, an ignition event is necessary. At this stage of the analysis, the location and time of the ignition leading to the rst stable ame propagation have to be determined. Ignition sources can be divided into random and intentional events (e.g. igniters). When igniters are taken into account in the analysis, the location and time of the rst ignition are well-dened by the development and expansion of the H2 /air cloud in the installation. As soon as an ignitable mixture ( 4% H2 ) reaches the rst active igniter, the ignition will be initiated. With correctly designed ignition systems, the ignition of the mixture will occur in the vicinity of the source shortly after the beginning of hydrogen discharge into the installation, which will contain only small amounts of hydrogen at this time. Without this intentional ignition, location and time of the rst ignition cannot be predicted mechanistically. A large number of potential ignition sources can be identied during an accident. In this case, the consequences of a random ignition have to be examined. Often, the worst-case ignition which cannot be excluded on mechanistic grounds has to be addressed. The determination of location and time of the rst ignition is important, since it represents the end of the non-reactive phase of the accident and denes the initial conditions of the now starting reactive phase.

9.3.2.2 Flame Acceleration After ignition, the combustion starts as a slow, quasi-laminar, premixed H2 /air deagration. It will propagate preferably into the direction with the highest ame speed. This means generally towards richer mixtures and into regions of high turbulence. This eect and also the self-induced turbulence which is generated due to the expansion of the burned gas behind the ame front may cause the transition from a slow laminar to a fast turbulent deagration. In order to derive general scaling laws for ame acceleration, FZK and the Kurchatov Institute Moscow (KI) performed joint test series in obstructed tubes.

350

W. Breitung

The tubes were geometrically similar with dierent diameters of 80, 174, 350, and 520 mm. Experiments in a large facility near Moscow provided data for a channel diameter of around 2250 mm. Optimum conditions for ame acceleration were created in these experiments (strong turbulence, repeated obstacles, no venting) to derive conservative criteria. H2 -air mixtures with additions of mono-, bi-, and triatomic inert gases were investigated (He, Ar, N2 , CO2 ) [8]. The systematic analysis of the measured data showed that the expansion ratio (= ratio of specic volumes [m3 /kg] of burned to unburned gas at constant pressure) is the most important mixture parameter, as it determines the ame propagation regime. The expansion ratio also represents the ratio of chemical energy in the unburned mixture to the initial thermal energy (Q/c pT0 ). Figure 9.8 displays the criterion for ame acceleration in H2 -air mixtures. The outcome of these experiments is that a borderline exists between tests with accelerating and with non-accelerating mixtures. The border correlates with 3.75 independently of the geometrical scale. This critical expansion ratio is valid for lean and rich hydrogen-air diluent mixtures under ambient conditions. In H2 -air mixtures at NTP, this limit corresponds to 10.5% H2 on the lean side and 74% H2 on the rich side. For rich mixtures, the ammability and ame acceleration limits coincide.

7 global quenching quenching-reignition FZK FZK choked flames quasi-detonation T ~ 300 K

Expansion ratio (= ub/b)

KI

FZK KI

fast flames

FZK FZK FZK / KI KI FZK FZK

KI

= 3.75

unstable flames slow flames

Fig. 9.8 Criterion for ame acceleration in H2 -air mixtures. Fast combustion regimes are only observed for mixtures with an expansion ratio above 3.75 (at an initial temperature T0 of about 300 K)

3 0.1

1.0

10.0

Tube diameter L / Laminar flame thickness x 103


different length scales L
FZK FZK-tube: 350 mm ~ ~ 0.1 mm FZK / KI RUT : 2400 mm KI Driver tube: 174 mm Torpedo tube : 520 mm

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

351

If this criterion is not fullled, a slow laminar combustion has to be expected, which may be modelled with a suitable code. In the case of the criterion showing the possibility of ame acceleration, the question is then whether the mixture predicted in the distribution calculation may even be capable of a transition to detonation.

9.3.2.3 Deagration-to-Detonation Transition (DDT) To predict whether such a transition may occur, the -criterion is applied as a decision tool in the FZK analysis methodology. The idea of DDT requiring some minimum size of the reactive mixture was rst proposed by researchers of the Kurchatov Institute (KI) and substantiated by a 1D numerical simulation [9]. A large number of joint KI-FZK experiments were subsequently performed in various facilities, partly with the participation of the French Nuclear Safety Authority (IRSN), to test this hypothesis. In addition, literature data on detonation onset conditions were collected and evaluated. Detailed analysis of these test data showed that in agreement with the numerical simulations a correlation exists between the occurrence of DDT and the geometrical size of the reacting gas mixture. Furthermore, it could be demonstrated with tests in scaled-down facilities that the detonation cell size of the mixture allows to consistently scale DDT in dierent mixtures and facilities. The nal result of the analysis is that a minimum volume is required for the onset of DDT, which can be expressed by D 7, where D is the characteristic size of the reactive gas cloud and is the average detonation cell size of the (generally non-uniform) gas mixture (Fig. 9.9). DDT is only possible, if this criterion is met.

Fig. 9.9 Criterion for detonation onset in H2 -air mixtures. D is the characteristic size of the combustible cloud (= V1/3 ), is the average detonation cell size of the mixture. For a given mixture, here one choosen with = 100 mm, the increase of the geometrical size D of the reacting cloud will cause a transition from the deagration regime to the detonation regime; in this example the critical size is near D = 700 mm

352

W. Breitung

Further details can be found elsewhere [10]. It is important to note that contrary to earlier literature data, DDT limits are scale-dependent. The larger the reactive cloud, the leaner mixtures can undergo a detonation transition. If the characteristic size is not sucient for DDT, a fast turbulent combustion has to be modelled, for which the COM3D code is used at FZK. Otherwise, the simulation of a detonation is necessary, e.g. with the code DET3D. The three transition criteria of 1. Flammability (inert to burnable mixture composition) 2. Flame acceleration (slow to fast deagration, -criterion) 3. Detonation onset (deagration to detonation, -criterion) are used to determine the appropriate combustion mode and select the corresponding numerical simulation tool for subsequent combustion modelling. Such transition criteria are necessary for the analysis of the hydrogen behaviour, since a direct numerical simulation of the transition process currently is not feasible. Much smaller time and length scales need to be resolved for such calculations, compared to the distribution and combustion simulation on a large scale. The mechanistic modelling of an ignition event, for instance, requires the resolution of the initial ignition kernel, including detailed chemistry, turbulence, and radiation losses. In the ame acceleration process the transition from laminar to turbulent combustion must be modelled. During the initiation of a detonation, the formation of a hot spot with strong ignition and its amplication in the surrounding mixture have to be resolved numerically. Direct simulation of these processes is currently hindered by a limited understanding of physical details and by insucient computational resources. It is important to note that the determination of the three empirical transition criteria used here only requires information about concentrations and geometrical dimensions of the burnable mixture formed during the accident. This information is available from the 3D distribution calculation. Therefore, the three criteria can be evaluated online during the distribution analysis to identify time and space domains with suciently extended and reactive mixtures (implemented in GASFLOW). The criteria can give early hints with respect to maximum possible combustion velocities and mechanical loads on the investigated structure without a full reactive combustion simulation. They are an easy way to check and optimise the eectiveness of the mitigation measures introduced at the beginning of the analysis. If, for instance, local detonations shall be excluded, it has to be shown that the -criterion is not fullled. The criteria described were implemented in an interactive program called GPcode which stands for Gas Properties [11, 12]. Figure 9.10 displays an example of the limits calculated for the dierent combustion regimes in H2 -air-steam mixtures at 400 K. The example relates to a characteristic size of the combustible cloud of D = 1 and 5 m, where D = (Volume)1/3 . As can be seen, the DDT limit is scaledependent. Indeed, a wider range of mixtures can undergo a detonation transition with increasing size of the system. In summary, the criteria described have two important functions in the analysis of the hydrogen behaviour. At rst, an early judgement as to the fastest possible

9 Safety Analysis of Hydrogen Vehicles and Infrastructure Fig. 9.10 Calculated limits for ammability, ame acceleration and detonation transition (DDT) in H2 -steam-air mixtures at 400 K. Output of GP-code [12]. The DDT limit is scale dependent, a wider range of mixtures can detonate in a larger system
0.8 0.7 Volume fraction hydrogen 0.6 0.5 0.4 0.3 0.2 0.1 Flammability limit 0.0 0.0 0.1 0.2 0.3 0.4 0.5 Volume fraction steam DDT limit Flame acceleration limit D = 1m D = 5m p0 = 0.1 MPa T0 = 400 K

353

0.6

combustion regime and the related pressure loads can be made without a detailed combustion analysis. Secondly, this knowledge allows for the selection of the appropriate numerical combustion program for further 3D analysis.

9.3.3 Third Phase: Combustion Simulation


The following 3D codes have been developed at the Hydrogen Safety Centre of the Forschungszentrum Karlsruhe to simulate the dierent hydrogen combustion regimes. They are used as follows: The FLAME3D and V3D codes for slow deagration (slow premixed deagration and non-premixed diusion ame) The COM3D code for fast premixed turbulent deagration The DET3D code for stable detonations in homogeneous H2 /air mixtures. In V3D a semi-implicit and in COM3D an explicit numerical scheme is used for the solution of the reactive Navier-Stokes equations. FLAME3D and DET3D use an explicit solver for the reactive Euler equations. Most of the combustion energy is transformed into heat, kinetic energy of the gas phase, and nally into pressure waves. Therefore, fast turbulent deagrations (modelled with COM3D) and stable detonations (modelled with DET3D) mainly lead to pressure and impulse loads, since the high combustion rates do not provide sucient time for signicant heat transfer to structures. All codes can model arbitrary geometrical boundary conditions and internal ow obstacles by using a structured mesh with cubic computational cells. A graphical user interface allows for an easy representation of complex 3D geometries with

354

W. Breitung

the aid of geometry elements (plates, cubes, spheres, etc.). An intensive validation process was performed for all physical models and numerical methods with the use of suitable experimental data on dierent geometry scales. This validation work was described and published recently [7].

9.3.4 Fourth Phase: Consequence Analysis


9.3.4.1 Mechanical and Thermal Loads The thermal and mechanical loads of the 3D combustion simulation, particularly the temperature and pressure histories at dierent positions in the installation, are recorded in the calculation. The time of the rst ignition determines which of these two load forms dominates. In both cases, the same combustion energy is released, but on very dierent time scales resulting in dierent heat release rates. In most cases, early ignition leads to low pressures, but high thermal loads due to stably standing diusion ames. On the contrary, late ignition of a hydrogen accumulation in an installation causes high pressure loads, but the temperature increase is negligible in large structures.

9.3.4.2 Structural Behaviour and Human Injury The calculated thermal and pressure loads represent the input for the last step of the analysis, which is the investigation of the structural behaviour and human injuries. Thermal loads due to stable diusion ames should not lead to structural failures caused by the degradation of sensitive components. Mechanical loads due to fast ames include pressure waves, impulses, and possibly the impact of debris generated by combustion pressure waves. If fast combustion processes cannot be ruled out with mitigation methods, further investigation of the structural behaviour is required to exclude structural failures. First-order estimates of the structural behaviour of building components under gas dynamic loads can be made with a single degree oscillator model. A numerical model was developed for this purpose [7] and an example of the application of this oscillator model can be found elsewhere [13]. If the single degree oscillator model indicates hazardous situations, more detailed 3D nite-element calculations should be applied to obtain more precise statements about the structural behaviour of the investigated installation.

9.3.5 Mitigation Measures


The primary goal of a safety analysis is to determine the risk which could arise from a given accident scenario. In case the risks are unacceptable, it is necessary to

9 Safety Analysis of Hydrogen Vehicles and Infrastructure Fig. 9.11 Possibilities for mitigation of accident consequences. The combination of such physically independent countermeasures can provide a robust defence- in- depth approach for a safe design

355

Exclude severe accident scenarios by design Limit hydrogen sources Support hydrogen mixing Exclude ignition sources Suppress flame acceleration avoid high H2 concentrations avoid confinement avoid flow obstacles - Avoid conditions for detonation on-set reduce H2 concentration limit cloud size - Limit damage by strong shelters

derive eective mitigation measures to reduce the hazard below a certain limit. The analysis methodology described above provides a good guideline on how accident consequences could be mitigated, as it covers the complete accident progression. Figure 9.11 displays a summary of the possibilities to interrupt or mitigate the progression of a hydrogen-related accident. The list of mitigation measures provided here follows the general sequence of events in an accident. The rst objective is to eliminate accidents with very high risks by changing the design of the installation. Once the design has reached the maximum possible degree of safety, the next line of defence is to limit the sources releasing hydrogen into the environment. The remaining hydrogen release must then be dealt with by the analysis. The next step to limit consequences is to exclude ignition sources as far as possible. If this cannot be achieved completely, the likelihood of fast ames should be decreased by reducing connement or by avoiding heavily obstructed geometries. In addition, favourable conditions for detonation onset should be avoided to the largest possible extent. Finally, as a last resort, a suciently strong shelter may be used to protect against severe accident consequences. A countermeasure generally is most ecient when it inuences accident progression in an early stage. The principle of defence-in-depth has proven to be very successful in other areas of energy technology. It tries to include in the design a series of safety barriers which are completely independent and rely on dierent physical principles. The combination of several of the measures described above can provide a robust defence-in-depth approach to a safe hydrogen-powered system.

9.4 Application in the Case of H2 Release in a Garage


Knowledge of the risks which may develop due to the release of hydrogen into conned spaces is of importance, especially in mobile applications. In the following sections, the analysis procedure described above shall therefore be applied to investigate the consequences of hydrogen release from a car parked in a residential garage.

Accident progression

356

W. Breitung

The scenario investigates the release at dierent rates of cold boil-o gas from a liquid hydrogen tank into initially stagnant air at ambient temperature in a simple rectangular garage [14]. It is assumed that the cryogenic LH2 tank leads to 170 g of hydrogen boil-o gas per day and that this amount is released in 5 venting intervals of 34 g each. The rst scenario considers this amount to be released in 10 s, while the release duration in the second scenario is 100 s. Table 9.2 summarises the dierent parameters for the investigated garage geometry and hydrogen release scenario. It should be noted here that the case of hydrogen penetration into conned compartments of the car is not addressed.

9.4.1 Mixture Generation


If the amount of 34 g H2 would be homogeneously distributed in the garage volume, then the resulting H2 concentration would be only 0.6 vol.%. This value is far below the ignition limit of 4 vol.% of H2 . Does such a hydrogen release represent any risk? The GASFLOW program [6] has been applied to simulate hydrogen release and subsequent mixing with air in the garage in the two scenarios above. Figure 9.12 displays the computed hydrogen distribution in the garage in the case of 34 g of hydrogen being released in 10 s. As can be seen, two separate plumes rise along two sides of the trunk within 7 s. Once the ow reaches the ceiling, a mushroom-shaped cloud forms. About 20 s after the beginning of hydrogen release, the cloud reaches its largest extension. The velocities in the rising plume are around 1 m/s. The momentum in the ow is redirected at the ceiling, causing an initially circular expansion of the burnable cloud and then reections at the nearby walls. These reections lead to a preferential ow along the ceiling towards the front of the car. Air is entrained by these convective gas motions, leading to a decrease of the combustible volume. Finally, the calculation predicts a stable stratication with most of the released hydrogen in a shallow layer underneath the ceiling. Figure 9.13 displays the computed H2 concentration in case of the slow release (0.34 g H2 /s). As can be seen, the calculation predicts a completely dierent distribution process. Here, only one thin, tube-like H2 -air plume extends from the source below the trunk to the ceiling. The plume is quasi-stationary with a slight pulsa-

Table 9.2 Garage geometry and hydrogen release scenarios investigated Case Geometry Volume (m3 ) 1 2 70.2 70.2 Vent Openings Two times 10 20 cm2 Hydrogen source Tot. Mass (g) 34 34 Duration (s) 10 100 H2 Rate (g/s) 3.40 0.34 Temper. (K) 22.3 22.3 Release Location Underneath the trunk

9 Safety Analysis of Hydrogen Vehicles and Infrastructure Fig. 9.12 Computed H2 -concentration eld in the garage for Case 1 in which 3.4 g H2 /s were released for 10 s. The shown isosurface depicts the region with burnable H2 -air mixtures (> 4% H2 )

357

7s

10 s

15 s

25 s

tion in the ow rate and diameter. Similarly to the rst scenario, a stably stratied hydrogen distribution develops in the garage, but with less gas motion along the ceiling. Shortly after the release of the total amount of gas (around 100 s), the combustible regions which are shown in red have nearly disappeared. This indicates that naturally mixing mechanisms exist, which can dilute this weak hydrogen source to unburnable mixtures. This example demonstrates the necessity of CFD modelling for reliable hydrogen safety investigations. A lumped parameter calculation of these two cases, treating the garage as one single volume would have resulted in a completely dierent result (0.6 vol.% H2 ) and an incorrect risk estimate.

9.4.2 Hazard Potential


For a risk assessment, it is necessary to know the hazard potential associated with the calculated hydrogen distributions in the garage. The criteria for the hazard potential described in Sect. 9.3.2 were evaluated from the computed time-dependent H2 -air distributions. Figure 9.14 displays the results for the three hazard levels, namely,

358 Fig. 9.13 Vertical cut through the computed H2 -concentration eld in the garage for Case 2, in which 0.34 g H2 /s were released for 100 s. The burnable region corresponds to the red volume (> 4% H2 )

W. Breitung

20 s

50 s

80 s

100 s

120 s

0.00

1.33 vol% H2

2.66

4.00

ammability, ame acceleration, and detonation transition. The results computed for a hydrogen release rate of 3.4 g H2 /s and 0.34 g H2 /s are given on the left and right hand side of Fig. 9.14, respectively. In the rst scenario, the high release rate of 3.4 g H2 /s leads to a maximum characteristic size dcc of the combustible cloud of around 1.6 m, which is equivalent to 4.1 m3 (dcc = Volume1/3 ). After the end of the release (10 s), the combustible cloud continues to grow due to the dilution of its enriched kernel by further entrainment of air. The maximum value for dcc is reached within 20 s. Once the maximum is reached, the cloud size decreases due to convective mixing by the

9 Safety Analysis of Hydrogen Vehicles and Infrastructure


180 160 140 120 100 80 60 40 20 0
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
3 2.5 2 1.5
0.02 0 0.08 0.06 0.04

359

CASE 1
90

CASE 2

dcc (cm)

80 70 60 50 40 30 20 10

dcc (cm)

10

20

30

40

50

60 3

0 0 0.12

50

100

150

200

250

Vfa (m )

0.10

Vfa (m )

10

20

30

40

50

60

0
1.4 1.2 1.0 0.8

50

100

150

200

250

D 7

D 7
no DDT

DDT
1 0.5 0 0 10 20 30 40 50 60

0.6 0.4

no DDT

0.2 0 0 50 100 150 200 250

Time,s
Fig. 9.14 Computed hazard parameters for a hydrogen release of 3.4 g H2 /s for 10 seconds (left) and a release of 0.34 g H2 /s for 100 s (right). Top: Characteristic dimension of the combustible cloud (4. . .75 vol% H2 ). Middle: Volume of cloud which could support spontaneous ame acceleration (10.5. . .75 vol% H2 ). Bottom: DDT index D/7 of cloud with H2 - concentrations between 10.5 and 75 vol%

momentum-induced gas ows. Finally, the volume of the combustible mixture approaches a constant value. The cloud which could support ame acceleration (middle left gure) grows directly with the start of hydrogen release and reaches a maximum volume around 0.7 m3 shortly after the termination of the release. The corresponding H2 inventory is 20 g. Thereafter, the cloud disappears in about 10 s. This growing and shrinking of the sensitive cloud result from the balance between the hydrogen source strength and the mixing processes. In the rst phase (< 10 s), the release dominates and in the second phase (> 10 s), the mixing dominates. Comparison of the time scales shows that the decay of the sensitive cloud ( 10.5%) feeds the growth of the combustible cloud ( 4%). The evaluation of the DDT potential shows even higher dynamics. The cloud reaches a detonable composition and size around 2 s after the hydrogen release starts. The most sensitive situation exists a few seconds later, during which hydrogen accumulates underneath the car. DDT possibility is predicted for a total duration

360

W. Breitung

of about 10 s (D/7 > 1). In analogy, the decay of the detonable cloud at > 8 s corresponds to the simultaneous growth of the cloud which is able to support ame acceleration (increase of Vfa in the gure above the DDT plot). In the second scenario with the slow release rate, dierent hazard parameters evolve. As can be seen, a much smaller combustible cloud is formed in this case (dcc 0.8 m, Vcc 0.5 m3 ). Moreover, about 20 s after beginning of the H2 release, a quasi-equilibrium is established between the source and the mixing processes, leading to a nearly constant cloud volume. After the termination of the release, the combustible cloud dissolves within 20 s, because the plume is cut o and regions with 4% of H2 or more cannot be sustained below the ceiling during the long release period. The cloud capable of supporting ame acceleration also is much smaller in this case compared to the rst scenario (about 0.1 m3 vs. 0.7 m3 ). Again a quasiequilibrium exists between sources and sinks. The cloud disappears within 10 s after the termination of hydrogen release. The maximum hydrogen inventory of the ame acceleration cloud was 2.7 g only. The evaluation of the DDT criterion D/7 shows that the size and hydrogen concentration of the enriched cloud are insucient in practice for a transition to detonation for the entire duration of the hydrogen release. Table 9.3 summarises the computed hazard parameters for the two scenarios investigated. From the computed results, it can be concluded that a strong dependence exists between the hydrogen release rate and the hazard parameters. In the case of a low hydrogen release rate, only a relatively small ammable cloud develops (< 0.5 m3 ). A combustible mixture is present in the rising plume only during the release period. Once the release stops, no combustible mixtures are left in the garage and no combustion hazard remains. At the large release rate, around 4 m3 of combustible mixture develop in the garage. As long as the source is active, an inner kernel of enriched mixture exists (up to 0.7 m3 ), which has the potential for sustaining supersonic combustion modes. However, these enriched regions dilute rapidly to more insensitive mixtures after the source is cut o. This process, on the other hand, supports the persistence of the deagration cloud ( 4% H2 ). At the end of the calculation, it shows a stably stratied distribution near the ceiling of the garage. This indicates that accidental ignition of such a mixture would lead to a substantial damage of the car and the garage. Therefore, active or passive countermeasures seem necessary in this case to control H2
Table 9.3 Summary of computed hazard parameters for hydrogen release in a garage [14] Case Flammable Cloud s (m3 ) 3.4 g H2 /s 0.34 g H2 /s 4.1 0.5 t (s) CAT = (50) RT + 15s = 115 Flame Acceleration Cloud s (m3 ) 0.7 0.1 m (g H2 ) 20 2.7 t (s) RT + 15s = 25 RT + 15s = 115 Detonable Cloud s (m3 ) 0.7 m (g H2 ) 20 t (s) 10

s = maximum size; m = maximum mass; t = time o f existence; CAT = complete analysis time; RT = release time

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

361

release rates in the range of few 0.1 grams per second. Such countermeasures need careful design calculations to prove the required eciency.

9.4.3 Combustion
The hazard parameters computed for the high hydrogen release rate of 3.4 g/s show that spontaneous ame acceleration and detonation onset could be possible. The GASFLOW calculations predict that almost 20 g of hydrogen can be involved in such a fast combustion event. The next step is then to analyse the consequences of such a local explosion in a residential garage. The corresponding experiments were performed at FZK in a closed test chamber representing the garage. Since the fraction of hydrogen burning in a fast mode can depend on many parameters, dierently sized combustion units were developed, which contained 2, 4, 8 or 16 g of hydrogen in a homogeneous stoichiometric H2 air mixture. These cubic combustion units contained layers of mesh wire to promote ame acceleration. They were ignited by a centrally located weak electric spark. Figure 9.15 displays the ame velocities measured inside the combustion unit for 8 and 16 g of hydrogen. Flame speeds typical of quasi-detonations were observed at the outer edge of the combustion unit. With this arrangement, local explosions with conservative overpressures were produced inside the test chamber. Figure 9.16 shows a photograph of the fast deagration obtained with the 8 g H2 combustion unit shortly after complete combustion of the H2 -air mixture. The pressure wave generated by the local explosion propagates into the surrounding air volume. The experiments were afterwards simulated with the COM3D code using the measured ame speeds inside the combustion unit as input data. The grid resolu-

Flame velocity , m/s

ignition
2g 2g 4g 4g 8g 8g 16g 16g

Radius of the combustion units,m

Fig. 9.15 Measured ame speeds in the combustion unit for 8 and 16 g of hydrogen. The combustion unit was a cube lled with stoichiometric H2 -air mixture, ignited by a weak electric spark in the middle (r=0). The ame speed reached quasi-detonation velocities at the outer edge of the cubes for 8 and 16 g of hydrogen inventory. The generated pressure waves have limiting, upper bound overpressures and impulses for the given hydrogen mass

362

W. Breitung

Fig. 9.16 Local explosion of 8 g of hydrogen in the FZK test chamber, simulating an accident scenario in a residential garage

tion used in the COM3D calculation was 3 cm. Figure 9.17 displays a vertical cut through the 3D pressure eld in the garage 7.7 ms after ignition of the 8 g H2 combustion unit located in the centre of the spherical pressure wave. At this time, the pressure wave emitted from the combustion unit into the surrounding air reaches the walls of the garage. A reected wave returns from the bottom of the test chamber. Figure 9.18 compares the measured and the computed overpressures for a position on the oor of the test chamber. As can be seen, the agreement is good between the experiment and the model at the beginning of the event. The dierences observed between experiment and simulation later are mainly due to the wall properties of the test chamber. They moved in the experiment, but were assumed rigid in the simulation.

Fig. 9.17 COM3D calculation of the fast deagration of 8 g of hydrogen in the test chamber. Shown is a vertical cut through the 3d pressure eld, 7.7 ms after ignition of the combustion unit which was located in the center of the spherical pressure wave

9 Safety Analysis of Hydrogen Vehicles and Infrastructure


0.3
Overpressure [bar]

363

Experiment Simulation

0.2 0.1 0.0 0.1


8g H2

0.2

0.19

0.21 0.20 Time [s]

0.22

Fig. 9.18 Comparison of measured and calculated overpressure for a position on the oor of the test chamber

It can be concluded that local pressure loads in complex conned geometries can be well predicted, if the ame velocity is correctly modelled.

9.4.4 Consequence Analysis


The output of the 3D combustion analysis allows to estimate the consequences of a local hydrogen explosion in a garage. Thermal and mechanical loads on structures and humans are of interest. The following discussion shall be limited to the eects of the pressure loads generated by a conned local explosion of 2 and 16 g of hydrogen, respectively.

9.4.4.1 Structural Response Figure 9.19 displays a comparison of the measured peak overpressures p+ and durations T + of the rst incident positive overpressure waves from conned local H2 explosions with damage thresholds of civilian buildings. The damage thresholds expressed in the form of p+ I + relations were taken from Baker et al. [15]. The positive impulse I + was converted into positive pressure duration T + by using the following equation: T + = 2 I + / p+ (9.4) Glass breakage depends on a number of parameters like glass thickness, glass area, and aspect ratio of the glass pane. In Fig. 9.19, two examples covering the typical overpressure range for glass breakage are plotted [15].

364
106
FZK Exp. 2g H2 FZK Exp. 16g H2
1,4m 1,4 2,4m

W. Breitung
T+=2I+/p+

p+

I+

Positive peak overpressure p+ [Pa]

T+

s Pa as 0 45 0 P 30

105
1m
1,4m

2,8m 5.6m 2,4m 4,0m

0 11

Partial demolition, 50%-75% of walls destroyed Major structural damage, some wrenched

s Pa

2m

104
glass breakage

4,0m 6,3m Area 1m2 4.7 mm thick

load bearing members fails Minor structural damage, wrenched joints and partitions

Area 3m2 4.7 mm thick

103 0.1

5m

2g H2

16g H2

Unconfined gaseous detonations

10 100 Duration of positive overpressure T+[ms]

1000

10000

Fig. 9.19 Comparison of measured blast wave parameters p+ and T+ from conned local H2 explosions to damage thresholds for civilian buildings

Three dierent degrees of damage to civilian masonry buildings are presented, ranging from minor structural damage to a signicant demolition of walls [15]. These lines represent the limit values for p+ and T + that have to be reached in order to cause the indicated degree of damage. Therefore, a pressure wave with parameters above these lines will even cause a higher degree of damage. A given damage may be caused either by a certain peak overpressure or by a certain overpressure impulse. In Fig. 9.19 the lines with a slope of 1 correspond to a constant impulse. The limiting impulse is given for each of the three levels of building damage (110, 300, and 450 Pas, resp.). The data points given in Fig. 9.19 indicate the p+ and T + values measured in the experiments with 2 g and 16 g of hydrogen. The distance from the ignition location of the stoichiometric H2 -air mixture to the pressure sensor location varies between 1.4 and 6.3 m. The peak overpressure decreases in general with increasing distance from the combustion unit. The spread in the measured p+ /T + points reects the eects of the connement and obstacles present in the test chamber (see Fig. 9.16). Indeed, obstacles can shadow pressure sensors (giving small p+ values). Furthermore, conning walls can lead to oblique or normal reection and focussing of pressure waves in 2D edges or 3D corners (giving high p+ values). Figure 9.19 also compares the data points measured under these conned and partially obstructed conditions with published correlations for unconned gaseous detonations. In [16] p+ /I + correlations were derived for unconned gaseous detonations of hydrogen-air mixtures. Here, these correlations are converted into a p+ /T + format. Results for 2 and 16 g of hydrogen are shown in Fig. 9.19. The

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

365

lines marked 1m, 2m, and 5m, resp., are lines for constant distance from the centre of the explosion. The peak overpressure increases with decreasing distance from the explosion origin and with increasing hydrogen mass. There is a quite good agreement in the general trends of the unconned load parameters with the data measured in the conned experiments. The deviations of the measured data points from the shown lines indicate the eect of the connement and obstructions on the blast wave parameters p+ and T + .

9.4.4.2 Human Injury The human organs which are most sensible to pressure waves are ear drums and lungs [17, 18]. Figure 9.20 displays a comparison of the blast wave parameters p+ and T + measured during the conned local H2 explosions described with the literature data for human injury thresholds of these organs. As can be seen, pressure waves with higher blast wave parameters p+ and T + cause increasing grades of injuries. In case of waves with a positive phase duration above 10 ms and a peak overpressure above 0.8 bar, for example, lung damage must be expected. Shorter waves need higher overpressures. The injury thresholds for ear drums were reviewed extensively by Richmond et al. [17]. Their statistical treatment of the experimental data resulted in the two bands shown in Fig. 9.20 for 1 and 50% probability of ear drum rupture. The width of the bands represents the scattering due to parameters like age or gender. With

106

1m 2m
1,4m 2,4m

p+ lung rupture threshold


NASA Baker
5.6m increasing grade of injury increasing grade of injury

I+

T+=2I+/p+

Positive peak overpressure p+ [Pa]

T+

105
1,4m

2,8m

2,4m

50 % eardrum rupture 4,0m 1 % eardrum rupture

4,0m

104

6,3m

5m

2g H2

16g H2

FZK Exp. 2g H2 FZK Exp. 16g H2

10

Unconfined gaseous detonations

0.1

100 10 Duration of positive overpressure T+ [ms]

1000

10000

Fig. 9.20 Comparison of measured blast wave parameters p+ and T+ from conned local H2 - explosions to human injury thresholds

366

W. Breitung

increasing p+ , the event of an ear drum rupture leads to increasing grades of ear injury, aecting also inner ear bones. The ear response is also dependent on the orientation of the incoming wave front with respect to the ear canal. The worst case is a normal incidence of the wave, which causes a higher reected overpressure. The same eect occurs, if the ear canal is oriented towards and located near a reecting surface. The p+ and T + values given in Fig. 9.20 correspond to the incident wave, assuming a nearby normal reecting surface. These data are conservative and appropriate for conned local explosions in small rooms like a garage. The data points given in Fig. 9.20 represent the p+ and T + values measured in the experiments with 2 and 16g of hydrogen. They indicate that local fast deagrations or detonations of 216 g of hydrogen can easily cause ear membrane rupture, but not lung damage. Signicantly larger amounts of hydrogen are needed for lung damage. The scaling laws for unconned detonations predict a doubling of T + for a 8-fold increase in the reacting hydrogen mass, which is due to the doubling of the (spherical) cloud radius. Therefore, if the hydrogen mass is increased to 128 g, the line for 16g H2 (8 16 = 128 g) is shifted by a factor of two to a higher T + value. Thus, the threshold for lung damage can be reached by an explosion of 128 g H2 at a distance of about 2 m.

9.4.5 Summary of Results of Garage Investigations


The transient release of hydrogen into a private garage was investigated, assuming that 34 g H2 were released with a mass ow rate of 0.34 and 3.4 g/s, respectively. The homogeneous distribution of this hydrogen mass would result in an inert mixture far below the ammability limit (0.6 vol.% of H2 ). In reality, complete mixing occurs after many hours only. The objectives were to analyse the risk and the possible consequences of such a release of hydrogen into a conned space. The 3D distribution calculations with GASFLOW identied the size, shape, composition, and time evolution of the reactive H2 -air clouds developing during and after the H2 release period (see Figs. 9.12 and 9.13). The knowledge of the hydrogen concentration distribution obtained from GASFLOW then allowed for a quantication of several risk parameters like size, hydrogen mass, and time of existence of burnable mixtures. These parameters are of importance, as the mixture could sustain a slow deagration, a fast turbulent deagration or even a detonation under inadvertent conditions (see Fig. 9.14). The dynamics of the reactive cloud evolution are governed by the balance between the H2 source strength and the physical mixing mechanism. The latter tends to dilute the pure hydrogen released from the source location. If the release rate is high, then the source dominates the mixing processes for the whole release period (10 s), leading to a large and continuously growing combustible cloud. In contrast, if the release rate is low, then the mixing dominates, leading to a small and practically constant cloud size during the whole release period (100 s).

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

367

The slow release rate (and total mass) creates only a low risk in terms of ammable cloud size, time of existence, and hydrogen mass involved (Table 9.3), whereas with the high release rate, a long-term potential for slow deagration is possible. Indeed, much of the injected hydrogen mass exists in the form of a stably stratied layer below the ceiling in the high-release scenario. As a consequence, fast turbulent deagration and detonation are possible during the release period itself (10 s) and around 15 s after the end of the H2 injection. A hydrogen mass of up to 20 g could generate fast combustion events. To evaluate the consequences of such local explosions in a garage, fast combustion experiments were performed in a closed test chamber at FZK using stoichiometric H2 -air mixtures containing 2, 4, 8 or 16 g of hydrogen. The used combustion units contained layers of mesh wire to promote ame acceleration and to create boundary (worst-case) pressure loads. The measured pressure histories were analysed with respect to peak overpressure and duration of the rst positive pressure wave ( p+ , T + ) and then compared with known thresholds for structural damage and human injuries. Fast combustion of 216 g of hydrogen could have the following consequences for the garage structure: window panes and other light garage components like e.g. the usual metal sheet door would break damage to a masonry wall would occur only, if the H2 explosion would take place near the wall (Fig. 9.19, local explosion with 16 g of H2 at 1 m distance) wooden framework constructions as often used in the USA for residential garages would be destroyed. A local explosion of 216 g in a garage would have the following impact on persons present in the garage: high probability of ear drum rupture no lung damage. It is clear from these results that H2 release rates of several g/s into a garage should be either prevented by adequate vehicle safety technology or mitigated by active or passive countermeasures installed in the garage. Such mitigation systems need careful design calculations to prove the required eciency. Mitigation measures are necessary also in case of the small leak rate investigated here (0.34 g/s), if the duration (and total H2 mass released) is more than 510 times larger.

9.5 Safety Issues for Hydrogen Vehicles and Infrastructure


In the previous chapter the example of hydrogen release from a car parked in a residential garage was investigated to demonstrate the course of a complete mechanistic safety analysis. The garage situation is only one accident scenario among many others. In a future hydrogen economy the spectrum of safety issues connected with the widespread use of hydrogen-powered cars will be signicant.

368

W. Breitung

Herein, an overview will be given as concerns the safety issues for hydrogen vehicles and infrastructure. The potential accident scenarios are divided into two classes, hazards resulting from vehicle operation and hazards connected with infrastructure (lling/repair/service stations).

9.5.1 Hazards Resulting from Vehicle Operation


A hydrogen risk develops only, if hydrogen is released from vehicle components containing hydrogen in liquid or gaseous form. By mixing with air, combustible H2 air clouds can be created and then pose a threat in case of ignition. The consequences of such a loss of hydrogen connement depend on many design details of the car. Figure 9.21 displays a simplied layout of a hydrogen car equipped with compressed gaseous hydrogen and a fuel cell. The hydrogen system can generally be divided into a high-pressure part and a low-pressure part. Both parts are separated by the pressure regulation valve (PRV). All the components of such a system are carefully selected and tested for their performance and compatibility with hydrogen. The test objectives vary from non-destructive to destructive, from single to endurance tests at the level of single components, sub-systems or the entire system. Generally recommended design guidelines for hydrogen cars are listed below [19]: Hydrogen leak prevention by the selection of appropriate materials and by minimising the number of anges and threaded connections Hydrogen leak detection Ignition prevention by minimising the sources for ignition. These guidelines provide for a reliable defence-in-depth concept for avoiding hydrogen-related incidents. Other good overviews of safety issues for hydrogen vehicles can be found in [20] and [21]. Nevertheless, the consequences of component failures, both in normal operation and during a collision, must be systematically evaluated to identify potentially unacceptable hazards and to prevent such scenarios by design countermeasures. In the following sections the dierent failure modes which can occur with a hydrogen/PEM fuel cell vehicle shall be considered [19]. They may be generated by

H2-Sensors Pressure relief valve GH2-Tank Region with high pressure (700 bar) Refueling connector Isolation valve Pressure regulation valve (PRV) Fuel cell system

Computer

Exhaust

Region with low pressure (15 bar)

Fig. 9.21 Simplied layout of the hydrogen containing components of a car using compressed gaseous hydrogen and a fuel cell

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

369

failures in the compressed hydrogen storage system, in the hydrogen delivery system, and/or in the fuel cell system. 1. Failure of the compressed hydrogen storage system a) Under normal operation Catastrophic rupture due to a manufacturing defect in the tank, defect caused by abusive handling of the tank, local destruction of epoxy resin in an area of the tank and stress rupture. Large hydrogen release due to faulty pressure relief valve tripping without cause, fault in tank wall. Slow hydrogen leak due to a defect in the tank, stress cracks in the tank liner from pressure cycling, faulty pressure relief device, and faulty coupling from tank to feed line. b) During a collision, potential failure modes for the storage cylinders are: Catastrophic rupture due to collision impact, puncture by a sharp object, external re with failure of pressure relief device to open. Large hydrogen release due to puncture by a sharp object, re-created hole in the tank, opening of pressure relief valve in a re (which is the purpose of the device). Slow hydrogen leak due to re-induced openings in the fuel line connection and impact-induced opening in the fuel line connection. 2. Failure of hydrogen delivery systems a) In normal operation: Large hydrogen release due to failure of fatigued connection in the highpressure line, simultaneous failure of check valve and quick disconnect shut-o. Slow hydrogen leak due to connection loosened by vibration, temperature, and/or pressure cycling, faulty solenoid or shut-o valve, faulty pressure regulator, or fatigue crack in the fuel line. b) Additional failure modes may arise due to collisions: Large hydrogen release due to shearing of the high-pressure fuel line, reinduced melting of solenoid valve, and rupture of pressure regulator valve. 3. Failure modes in the fuel cell system The fuel cell system consists of the fuel cell stack and additional auxiliary components, such as air compressors, gas humidication equipment, heat exchangers, and others. During normal operation, a hydrogen leak may develop and lead to combustible mixtures in distribution manifolds, humidication system, heat exchanger, hydrogen recirculation compressor, car compartments or the exhaust system. Also purging the fuel cell may lead to hydrogen and to combustible H2 -air mixtures release into the environment. In case of a collision, no additional risk is expected from

370

W. Breitung

the fuel cell, since the solenoid valves will have shut o the hydrogen ow before it can reach the fuel cell [19]. The fuel cell system, however, should be designed to minimise the risk of hydrogen ignition. Over the past decade, car manufacturers carefully addressed these issues by appropriate testing and research. An example is the activities carried out by BMW. They started in the 1990s on the component level [22] and meanwhile have produced convincing safety results for the complete car system [23]. The consequences of the failure modes described above will depend strongly on the degree of connement of the released hydrogen gas. Scenarios with a high degree of connement of the produced H2 -air cloud have a much higher potential for fast combustion modes than those with H2 release into an open environment. In mobile applications all accident situations with restricted possibilities of hydrogen dispersion therefore are of special concern: Parking in a residential or public garage Collision-induced H2 release in a tunnel Hydrogen leak from a car in a ferry Accidents underneath bridges or in narrow street canyons

Other topics which may require special safety investigations in the future are vandalism and neglect/misuse of hydrogen vehicles and equipment.

9.5.2 Safety Issues for Refuelling Stations


The most recent and comprehensive European eort to analyse the safety of hydrogen refuelling stations was undertaken within the EU-funded HyApproval project [24]. This project combined expertise from petrochemical industry, suppliers of hydrogen lling station equipment, research organisations, and licensing authorities. In the following sections, a summary shall be given as concerns the main open questions and issues connected with the installation of hydrogen refuelling stations in urban areas. The nal results of the HyApproval project will be published in early 2008.

9.5.2.1 Accident Scenarios Figure 9.22 displays a schematic diagram of the main components of a hydrogen lling station with supplies of compressed gaseous hydrogen (CGH2 ) and liquid hydrogen (LH2 ). In this design a reformer and an electrolyser are included for onsite production of gaseous hydrogen. The cost scaling and the technical complexity of LH2 production favour a centralised large LH2 plant and the delivery of LH2 by trailers to the lling station. For each sub-system of the lling station, potential accident scenarios were identied at the beginning of the project. Possible leak sizes were classied into very small, small, medium, and large size, which corresponds to leakage from the gland

9 Safety Analysis of Hydrogen Vehicles and Infrastructure


Reformer Elektrolyser CGH2- tanker delivery H2compressors CGH2storage CGH2- supply

371

CGH2dispensor

LH2-tanker delivery

LH2storage

LH2pumps

LH2dispenser

LH2- supply

Fig. 9.22 Main components of a hydrogen lling station with supply lines for compressed gaseous hydrogen (CGH2 ) and liquid hydrogen (LH2 )

packing of valves, crack or pinhole, partial opening of a valve, and rupture of a storage vessel [25, 26]. Hazards and accident scenarios were identied for each of the lling station components. Some examples are given below concerning the line delivering compressed gaseous hydrogen. 1. Reformer inside closed container: leakage or rupture of the natural gas feed line, rupture of the reforming tube, pipe rupture due to hydrogen-induced embrittlement, rupture of a hydrogen pipe, leakage or burst of the H2 line to the compressor.

The main safety barriers which can be used to protect structures and personnel are ventilation inside the container, detectors for natural gas and hydrogen combined with an isolation system, venting surfaces on the container, restricted access to the container, regular inspection of the reformer installation.

2. Electrolyser inside closed container: Accident scenarios are related to the hazardous substances of lye and hydrogen. Lye leak through cells due to overpressure or gasket failure, leading to explosion or jet ame inside the container, lye leakage from pipes, oxygen leak inside the container, leading to re enhancement, hydrogen leak inside the container, leading to explosion or jet re, large H2 ingress into container by backow of H2 from the storage vessel.

372

W. Breitung

The main safety barriers in this case are preventing a mixture of H2 and O2 inside the electrolyser, H2 and O2 detectors with an isolation system for up- and down-stream direction, venting surfaces on the container, restricted access to the container, regular inspection of installation. 3. Compressor inside container: Hydrogen is pressurised by a compressor from about 10 to several 100 bar. Since the compressor is connected to high-pressure buer tanks, leakage and back-ow from the buer tanks are important scenarios. Further unwanted events are excessive pressure at the compressor outlet, compressed air ingress into the hydrogen line, leakage of hydrogen due to vibrations. As possible safety barriers, pressure control, vibration control, hydrogen detection, and forced ventilation were identied. 4. Buer storage in open air: The buer storage contains the major hydrogen inventory handled in the refuelling station. The main hazardous phenomena are hydrogen leaks with subsequent explosion or jet re and bursting of hydrogen pressure vessels due to mechanical or thermal loads. Due to the high pressures and volumes of H2 bottles, a burst would also create a large mechanical energy release. No sucient data are currently available to predict whether bursting of one bottle would trigger the failure of others bottles in the bundle. The simulation of hydrogen release from a high-pressure storage system can be found elsewhere [27]. The main safety barriers against such scenarios are the ability to limit hydrogen leak rates, to isolate a leaky storage bottle, and to discharge the hydrogen inventory in a safe way. Moreover, the bottles must be safely vented in case of re. 5. Dispenser in open air: The dispenser consists of the refuelling unit and the dispenser hose. The dispenser is generally covered by a weather protection shield. The use of a exible hose and the connection/disconnection actions with the car signicantly increase the probabilities of a hydrogen leak or a line rupture. One scenario of many others is the leakage at the lling nozzle due to the deterioration of the isolation valve [25]. Other unwanted events are hose rupture during/after refuelling and backow from the vehicle tank. The dispenser scenarios are critical for the whole safety evaluation of the lling station, because customers are involved and located close to the release location. This scenario was investigated experimentally by Tanaka et al. [28] and Takeno et al. [29]. Results from tests with premixed H2 -air clouds and H2 jet releases can be found in [30]. Possible precautious and safety barriers are hydrogen leak detection before connection, good nozzle and hose maintenance. Furthermore, an automatic lling procedure can be envisaged, where the customer is not involved and where an emergency shutdown procedure is implemented with well-dened actuation criteria.

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

373

6. CGH2 tanker delivery in open air: Instead of on-site production, CGH2 could also be delivered to the lling station by a CGH2 cylinder truck. This operation involves the transfer of large quantities of hydrogen at high pressures. For a 700 bar lling station, the delivery pressure must be well above 1000 bar to obtain a reasonable discharge time and discharge fraction. Important accident scenarios are the leakage of hydrogen from the trailer hose, hose failure during relling, or leakage after decoupling from the station side. Important safety barriers in this case are hydrogen leak detection before, during, and after tanker connection, loading of the storage tanks in a separately protected area, and emergency shut-down procedures. 9.5.2.2 Risk Assessment After identication of possible initiating events and accident scenarios, the next task is to quantify the risk connected with each event and to identify the dominating risk contributors. The EU project HyApproval uses two methodologies which are probabilistic risk assessment and computational uid dynamics (CFD) calculations. These methods complement each other by providing information on dierent levels of detail. The rst stage of the standard risk assessment approach is hazard identication. Its objective is to identify the major hazards that could contribute signicantly to the risk caused by the refuelling installation. Then, the frequencies and consequences of the hazards are calculated. These are combined (often by a multiplication) to derive a measure of the risk of each hazard. Next, the obtained risk is compared with an acceptance criterion to decide whether this risk is acceptable or not. If not, riskreducing countermeasures by both technical and operational improvements must be investigated. For the leading risk contributors, it is necessary to obtain more detailed information on re and explosion hazards which may aect station personnel or customers mainly by heat radiation and pressure waves. In the EU project HyApproval credible accident scenarios were selected for CFD simulations, which are based on experience gained by gas industry [31]. These include CGH2 dispenser failures at 35 and 70 MPa, CGH2 tanker discharge hose failure at 25 MPa, and LH2 dispenser failure at 38 bar. A complete list of the selected accident scenarios can be found in [31]. It should be noted that some of the scenarios given in the Chap. 19.1 are considered to be too improbable like the catastrophic failure of a CGH2 buer storage vessel. Therefore, no CFD simulation will be made for such a case.

9.6 Outlook
Additional research activities will be needed in order to improve the knowledge and, in turn, mitigate the above potential hazards from the use of hydrogen in vehicle operation and in infrastructure installations. General issues which are largely

374

W. Breitung

independent of the specic vehicle design will be addressed. It would be very convincing, if these topics could be resolved by the automobile community in a joint effort to generate internationally accepted and harmonised hydrogen safety standards and procedures. Issues related to the development of safe fuel cell cars can be found in [32]. With respect to vehicle safety, the following research topics will deserve closer attention in the future: safe venting of compressed hydrogen gas cylinders (350 and 700 bar), optimum arrangement of H2 storage vessels in the vehicle, re safety of hydrogen-powered vehicles with the primary goal to prevent bursting of the high-pressure hydrogen system, guidelines for re ghters in case of re or accident, optimum number and location of hydrogen detectors, safety concept in case of a hydrogen leak detection in a running car, tolerable H2 leak rates in the vehicle for dierent operating conditions, including a parked car, optimum position and activation criteria for pressure relief devices on the H2 tank, procedures to prevent penetration of hydrogen into the passenger compartment, eectiveness of forced ventilation for reducing local H2 concentrations in sensitive car areas, maximum possible reduction of ignition sources, development of standardised safety test procedures for new solid storage materials, such as nanocrystalline powders. Safety investigations for hydrogen infrastructure installations like maintenance, service, repair, and lling stations should address the following research topics: determination of tolerable H2 releases during vehicle repair, which pose no risk to the personnel, design of eective and low-cost ventilation systems, CFD analysis of leaking hydrogen scenarios, including complex surroundings near the vehicle, extension of the investigations described in [33], control of ignition sources, in case of lling stations, the issues of protecting walls and safety distances need to be investigated. More basic research is also needed to further improve the current knowledge of hydrogen properties: measurement of ignitable space regions, given a certain leak size, shape, and mass ow rate, an extension of the work described in [34], systematic investigation of active and passive safety systems, e.g. ventilators, catalytic recombiners, or ame arrestors, modelling of ignition processes under realistic boundary conditions, investigation of diusion ame stability after ignition (limits for lift-o and extinction),

9 Safety Analysis of Hydrogen Vehicles and Infrastructure

375

criteria for ame acceleration and detonation onset in H2 -air mixtures with concentration gradients and partial connement (Note: the criteria described in Sect. 9.3.2 are valid for homogeneous and fully conned mixtures; they are, hence, very conservative with respect to practical accident conditions in mobile applications and should be extended to more prototypic conditions).

References
1. M. Hattwig, H. Steen (Eds.), Handbook of Explosion Prevention and Protection, Wiley-VCH Verlag Weinheim, 2004, ISBN 3-527-30718-4. 2. A. Veser, J. Grune, G. Stern, W. Breitung, S. Dorofeev, Flame acceleration in a vented explosion tube, 19th Int. Coll. On the Dynamics of Explosions and Reactive Systems (ICDERS), Hakone, July 27August 1, 2003, Paper 183, Proc. on ROM. 3. A.M. Kreiser, G. Fr ohlich, M. Eichert, A. Schatz, Analyse von St orf allen mit Wassersto in bisherigen Anwendungsbereichen mit besonderer Ber ucksichtigung von LH2, Report IKE 2-116, Universit at Stuttgart (April 1994). 4. R.G. Zalosh, Comparative Analysis of Hydrogen Fire and Explosion Incidents, Report COO4442-1, Factory Mutual Research Corporation, Norwood, Massachusetts, USA (December 1977). 5. W. Breitung, U. Bielert, G. Necker, A. Veser, F.-J. Wetzel, K. Pehr, Numerical Simulation and Safety Evaluation of Tunnel Accidents with a Hydrogen Powered Vehicle, Proc. of the 13th World Hydrogen Energy Conference, June 1215, 2000, Beijing, China, Vol. 2, pp. 11751181. 6. J.R. Travis, P. Royl, R. Redlinger, G. Necker, J.W. Spore, L.L. Lam, T.L. Wilson, B.D. Nichols, C. M uller, GASFLOW-II: A Three-Dimensional Finite-Volume Fluid Dynamics Code for Calculating the Transport, Mixing, and Combustion of Flammable Gases and Aerosols in Geometrically Complex Domains, Vol. 1, Theory and Comp. Manual, Vol. 2, Users Manual, Reports FZKA-5994, LA-13357-MS (1998). 7. W. Breitung, W. Baumann, U. Bielert, B. Burgeth, S. Dorofeev, B. Kaup, A. Kotchourko, G. Necker, R. Redlinger, P. Royl, J. Staringer, G. Stern, J.R. Travis, A. Veser, Z. Xu, Innovative Methoden zur Analyse und Kontrolle des Wasserstoverhaltens bei Kernschmelzunf allen, Report FZKA 7085, Forschungszentrum Karlsruhe Germany (February 2005), available in full text under www.fzk.de/hbm (free search). 8. S.B. Dorofeev, M. Kuznetsov, V. Alekseev, A. Emenko, W. Breitung, Evaluation of limits for eective ame acceleration in hydrogen mixtures, Journal of Loss Prevention in the Process Industries 14, 2001, pp. 583589. 9. S.B. Dorofeev, A. Kotchourko, S.B. Chaivanov, Detonation Onset Conditions in Spatially Non-Uniform Combustible Mixtures, Proc. of the 6th Int. Symposium on Loss Prevention and Safety Promotion in the Process Industries, 1989, Oslo, Vol. 4, pp. 2221. 10. S.B. Dorofeev, V.P. Sidorov, M.S. Kuznetsov, I.D. Matsukov, V.I. Alekseev, Eect of scale on the onset of detonations, Shock Waves 10, 2000, pp. 137149. 11. A. Gavrikov, A. Lelyakin, W. Breitung, S. Dorofeev, Calculation of Detonation Cell Sizes for Hydrogen-Air-Steam Mixtures at ITER Typical Pressures, Forschungszentrum Karlsruhe, Germany, Report FZKA 6812 (2003). 12. M. Kuznetsov, A. Lelyakin, S. Dorofeev, Flammability Limits for Nuclear Safety Applications, Report of Kurchatov Institute, prepared for FZK-IKET (1999), and A. Lelyakin, A. Gavrikov, S. Dorofeev. A Computer Code Package to Calculate Explosion Properties of Hydrogen Combustibles, Report of Kurchatov Institute, prepared for FZK-IKET, Moscow (2001).

376

W. Breitung

13. W. Breitung, Analysis Methodology for Hydrogen Behaviour in Accident Scenarios, Proc. Int. Conf. on Hydrogen Safety, September 810, 2006, Pisa, Italy. 14. W. Breitung, G. Necker, B. Kaup, A. Veser, Numerical Simulation of Hydrogen Release in a Private Garage, Proc. of the 4th Int. Symposium on Hydrogen Power HYPOTHESIS IV, September 914, 2001, Stralsund, Germany, Vol. 2, p. 368. 15. W.E. Baker, P.A. Cox, P.S. Westine, J.J. Kalesz, R.A. Strehlow, Explosion Hazards and Evaluation, Elsevier Publ. Co., Amsterdam, 1983, p. 596. 16. S.B. Dorofeev, Blast Eects of Conned and Unconned Explosions, Proc. 20th Symposium (Int.) on Shock Waves, July 1995, Pasadena, California, USA, pp. 7786. 17. D.R. Richmond, E.R. Fletcher, J.T. Yelverton, Y.Y. Phillips, Physical correlates of ear drum rupture, Annals of Otology, Rhinology & Laryngology 98, 1989, pp. 3541. 18. NASA Oce of Safety and Mission Assurance, Safety Standard for Hydrogen and Hydrogen Systems, Report NSS 1740.16 (1997). 19. C.E. Thomas, Direct-Hydrogen-Fueled Proton-Exchange-Membrane Fuel Cell System for Transport Applications, Report by Ford Motor Comp. Dearborn, MI, USA, for US-DOE, Report DOE/CE/50389-502 (May 1997). 20. J.T. Ringland, Safety Issues for Hydrogen-Powered Vehicles, Report SAND94-8226, Sandia Natl. Laboratories Livermore Cal., USA (March 1994). 21. L.C. Cadwallader, J.S. Herring, Safety Issues With Hydrogen as a Vehicle Fuel, Report INEEL/EXT-99-00522 (September 1999). 22. K. Pehr, Experimental Examinations on the Worst Case Behaviour of LH2/LNG Tanks for Passenger Cars, Proc. 11th World Hydrogen Energy Conference, June 2328, 1996, Stuttgart, Germany, Vol. 3, p. 2169. 23. J.-M. Vernier, C. M uller, S. F urst, Safety Measures for Hydrogen Vehicles with Liquid Storage, Proc. of 16th World Hydrogen Energy Conf., June 1316, 2006, Lyon, France. 24. HyApproval internet page: www.hyapproval.org. 25. S. Lim, L. Perrette, Risk Assessments and Accident Simulations as per Matrix Table, Report by INERIS, HyApproval Deliverable WP4, ST3 (October 2006), www.hyapproval.org. 26. L. Perrette, S. Lim, Proposed List of Scenarios for the Modelling Task, Report INERIS HyApproval Deliverable WP4, 4.X (June 2006), www.hyapproval.org. 27. B. Angers, P. B` enard, A. Hourri, P. Tessier, J. Perrin, Simulations of Hydrogen Releases from High Pressure Storage Systems, Proc. of 16th World Hydrogen Energy Conf., June 1316, 2006, Lyon, France. 28. T. Tanaka, T. Azuma, J.A. Evans, P.M. Cronin, D.M. Johnson, P. Cleaver, Experimental Study of Hydrogen Explosions in a Full-Scale Hydrogen Filling Station, Proc. Int. Conf. on Hydrogen Safety, September 810, 2005, Pisa, Italy. 29. K. Takeno, K. Okabayashi, A. Kouchi, T. Nonaka, K. Hashiguchi, K. Chitose, Phenomena of Dispersion and Explosion of Highly Pressurized Hydrogen, Proc. Int. Conf. on Hydrogen Safety, September 810, 2005, Pisa, Italy. 30. L.C. Shirvill, P. Roberts, C.J. Butler, T.A. Roberts, M. Royle, Characterization of the Hazards from Jet Releases of Hydrogen, Proc. Int. Conf. on Hydrogen Safety, September 810, 2005, Pisa, Italy. 31. Report by Shell Hydrogen, Establishment of Best Practices for Safety; HyApproval Deliverable 4.2 (November 2006), www.hyapproval.org. 32. L. Perrette, H. Paill` ere, G. Joncquet, Presentation of the French National Project DRIVE, Proc. of 16th World Hydrogen Energy Conference, June1316, 2006, Lyon, France. 33. Technical report for California Fuel Cell Partnership (CaFCP), Support Facilities for Hydrogen-fueled Vehicles. Conceptual Design and Cost Analysis Study (July 2004), info@cafcp.org. 34. M. Swain, Codes and Standards Analysis DE-FC36-00GO 10606, A007, Final Technical Report 04/15/0304/14/04, Univ. of Miami, Coral Gables, Florida, USA. General recommendation for further literature: Extensive literature on all aspects of hydrogen safety can be found in the US-DOE database: www.hydrogen.energy.gov/biblio database.html.